/Tech5h ago

AI safety researcher Chris Painter disputes Erik Brynjolfsson's claim that deployed LLMs actively use 'neuralese

Painter notes the concept refers to undocumented latent-vector reasoning.

10385147K
Chris Painter
Erik Brynjolfsson#496
Original post
Erik Brynjolfsson@erikbryn#496inTech

Neuralese: The highly compressed, internal language increasingly used by some LLMs. Unlike English, Neuralese can't be understood by humans.

11:40 AM · Jun 11, 2026 · 6.7K Views
Sentiment
Sentiment building, check back later.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS420LIKES6REPLIES1
Chris Painter@ChrisPainterYup

Just to clarify, I would not say neuralese is “increasingly used by some LLMs”. I generally understand this term to mean “recurrent latent-vector reasoning, not implemented via an output chain of thought”. There’s some evidence models can reason without using the CoT, but “neuralese” has not been publicly documented as being used in publicly deployed models.

METR@METR_Evals

Fact 5: Agents seemed to rely on natural language to reason through the hardest tasks, and no shared model had an architecture that allowed for opaque recurrence. Frontier performance was much worse when agents were unable to “reason out loud” or had to hide what they were doing.

2hViews 420Likes 6Bookmarks 0